Room equalization through inverse delayed and attenuated bass signals

[neRok]

The method I mentioned at the end of my last post about creating cancel tones in Audacity is proving fruitful. Because what I've come to realise is that VBA as we are doing it boils down to adding up impulse responses at the listening position. Because presuming that the speaker can output the cancel tone (which means ignoring the changes that can happen when you mix a cancel tone in to the source), that's what will happen (summed IR's), because the cancel tone is basically just another tone that will travel around the room in exactly the same way as the main tone, and we will experience its effects at the LP. So adding IR's together is basically what we are doing - the VBA wave being different due to any amplitude or phase changes desired.

This seems an important step in designing VBA like DL-ART is doing because it also allows simulating one speaker or subwoofer countering another. If L is cancelling R, L cancel impulse at the LP will be slightly different than if R output that same cancel signal.

Additionally I came to realised that a full cycle tone is a half cycle IR (eg start with a crest), and then the next half cycle (the valley) is just an inverted and delayed IR too. Add the 2 half cycle IR's together and it's the same as doing the IR of the full cycle tone. So I've taken to using half cycle only in the very first stages of trial and error, because it seems less "noisy".

So just to set the baseline for the discussion, below is 33Hz 1x full cycle tests.

The 1st row isn't the test tone, I just put it there to see the subtle phase shift on the following results.
The 2nd row is the tone passed through the in room IR, and then normalised (made to full volume). The orange lines are the rough shape of a pure tone, and everything after the orange lines is un-decayed tone.
The 3rd row is the tone passed through the VBA IR, then the in room IR. I amplified this tone so that the second crest at ~30ms was the same amplitude as the previous result. This means the 3rd crest is quieter, because the VBA has volume shifted it. It has reduced the 4th crest but possibly by too much, improved the 5th crests, and reduced the decay thus improving it.

Next I begin my tests with a half cycle tone put through the in room IR (row 1). Then I duplicated it, and modified the duplicate as I wanted (row 2). I delayed it so that a peak of choice aligned with another peak of choice, and then row2 was inverted as required so that they cancelled. Then I mixed the 2 together to get row 3.
The first attempt was not very successful. Lots of decay still.


The second attempt I adjusted the delay and inverted it. Some of the decay side is improved. Where the cursor line is (~150msec) you can see the IR's added up to make decay noise. But I could see that if I shifted the cancel tone to the left slightly, they would sum - to +.


Third attempt improved the cursor position, but the rest isn't great.


Last attempt for this post, and I delayed it quite a bit. This seems a good result! The cancellation in the decay area seems good.


So I took that attempt and fleshed it out fully (with a full cycle tone), so that I could compare it with the previous VBA. So I just manually place the 2 tones, mix them, then IR the mix. Here it is;

Next screenshot is the same thing but normalised. What would the actual in room SPL be, I'm not sure? Maybe the first or second peak should be volume matched. So normalising probably isn't proper, but also maybe it is when volume matched in room.


So that's a decent result! The decay seems improved versus the old VBA example. The other big difference to me is that the 4th crest is a bit stronger, which I think is better, because in the first screenshot you can see it would fall below the orange line.

The VBA as OCA did it would probably achieve the same result if it was delayed/amplified the same, but this method seems easier to line everything and see the decay, instead of doing it in REW and only seeing the SPL result. But also, I haven't even got to trialling multiple VBA tones of different frequencies yet, with 1 of them possibly being a weak inverted tone ahead of the main tone.

Anyway, this is not a production ready workflow, but I think it shows promise, because this is letting me work out the optimal delay and amplitude of the VBA tone. AND this workflow will be good for speaker A cancelling speaker B.

 

[tjcinnamon]

The method I mentioned at the end of my last post about creating cancel tones in Audacity is proving fruitful. Because what I've come to realise is that VBA as we are doing it boils down to adding up impulse responses at the listening position. Because presuming that the speaker can output the cancel tone (which means ignoring the changes that can happen when you mix a cancel tone in to the source), that's what will happen (summed IR's), because the cancel tone is basically just another tone that will travel around the room in exactly the same way as the main tone, and we will experience its effects at the LP. So adding IR's together is basically what we are doing - the VBA wave being different due to any amplitude or phase changes desired.

This seems an important step in designing VBA like DL-ART is doing because it also allows simulating one speaker or subwoofer countering another. If L is cancelling R, L cancel impulse at the LP will be slightly different than if R output that same cancel signal.

Additionally I came to realised that a full cycle tone is a half cycle IR (eg start with a crest), and then the next half cycle (the valley) is just an inverted and delayed IR too. Add the 2 half cycle IR's together and it's the same as doing the IR of the full cycle tone. So I've taken to using half cycle only in the very first stages of trial and error, because it seems less "noisy".

So just to set the baseline for the discussion, below is 33Hz 1x full cycle tests.
View attachment 303897
The 1st row isn't the test tone, I just put it there to see the subtle phase shift on the following results.
The 2nd row is the tone passed through the in room IR, and then normalised (made to full volume). The orange lines are the rough shape of a pure tone, and everything after the orange lines is un-decayed tone.
The 3rd row is the tone passed through the VBA IR, then the in room IR. I amplified this tone so that the second crest at ~30ms was the same amplitude as the previous result. This means the 3rd crest is quieter, because the VBA has volume shifted it. It has reduced the 4th crest but possibly by too much, improved the 5th crests, and reduced the decay thus improving it.

Next I begin my tests with a half cycle tone put through the in room IR (row 1). Then I duplicated it, and modified the duplicate as I wanted (row 2). I delayed it so that a peak of choice aligned with another peak of choice, and then row2 was inverted as required so that they cancelled. Then I mixed the 2 together to get row 3.
The first attempt was not very successful. Lots of decay still.
View attachment 303898

The second attempt I adjusted the delay and inverted it. Some of the decay side is improved. Where the cursor line is (~150msec) you can see the IR's added up to make decay noise. But I could see that if I shifted the cancel tone to the left slightly, they would sum - to +.
View attachment 303899

Third attempt improved the cursor position, but the rest isn't great.
View attachment 303900

Last attempt for this post, and I delayed it quite a bit. This seems a good result! The cancellation in the decay area seems good.
View attachment 303903

So I took that attempt and fleshed it out fully (with a full cycle tone), so that I could compare it with the previous VBA. So I just manually place the 2 tones, mix them, then IR the mix. Here it is;
View attachment 303902
Next screenshot is the same thing but normalised. What would the actual in room SPL be, I'm not sure? Maybe the first or second peak should be volume matched. So normalising probably isn't proper, but also maybe it is when volume matched in room.
View attachment 303901

So that's a decent result! The decay seems improved versus the old VBA example. The other big difference to me is that the 4th crest is a bit stronger, which I think is better, because in the first screenshot you can see it would fall below the orange line.

The VBA as OCA did it would probably achieve the same result if it was delayed/amplified the same, but this method seems easier to line everything and see the decay, instead of doing it in REW and only seeing the SPL result. But also, I haven't even got to trialling multiple VBA tones of different frequencies yet, with 1 of them possibly being a weak inverted tone ahead of the main tone.

Anyway, this is not a production ready workflow, but I think it shows promise, because this is letting me work out the optimal delay and amplitude of the VBA tone. AND this workflow will be good for speaker A cancelling speaker B.

This is a great development. Looking forward to it becoming further refined.

 

[OCA]

this is not a production ready workflow, but I think it shows promise

Thanks for all that information. Possibility to be able to use another channel rocks! I hope you will also share instructions for people unfamiliar with Audacity like myself

 

[Tim Link]

I turned my room sideways last night. I've been playing with it all day today. The horns are now too far apart to work together to create a plane wave. I can't get an inverse delayed signal to do anything good now. My peaks and dips look different. I was able to put in an FIR filter, non minimum phase, and get a practically perfect looking impulse response and frequency response from 10Hz up through the rest of the bass. It doesn't sound right though. There's something weird about it. The bass was better with both speakers on a small wall. The imaging and room useability is better this way. Since I can't figure out how to not make FIR filters sound weird, minimum phase or not, I've gone back to good old parametric EQ! I've not given up on FIR, but I need to learn more, maybe get tools more flexible than REW, or or learn how to use it better.

One thing I can get is flat response down to 10 Hz, whatever good that does.

 

[neRok]

I hope you will also share instructions for people unfamiliar with Audacity like myself

I've just uploaded a quick video now showing the steps I use;

I also show at the end how (tone+vba)>convolved is the same is (tone>convolved)+(vba>convolved), because like I said before, it's basically just 2 tones going out in to the room.

Just a quick comment on Audacity - if you get 64bit Audacity (which you should), it can only use 64bit plugins (which many are). The free convolver I've settled on is MConvolutionEZ.

Edit: It's probably a good idea to mute Audacity in Windows/OS, because it's quite easy to accidentally "start playback" and you wouldn't want to blast some random tones that are possibly clipped out your speakers.

 

[neRok]

I've been having a few goes at making an IR to match the manual VBA I came up with the other day, which was 60ms delay, inverted, and -6dB gain. I had to have a few goes because you can do the steps in different orders in REW, and I'm not sure how important the order is? Actually many tests gave the similar results in Audacity, so it might not be that important. But also, the cancellation impulse in Audacity wasn't "clean" and looked more like a "wave packet", and it was further delayed. I'm not sure why? But I rolled with it and adjusted it further so that the main peak of the VBA IR lined up with my manual test.

So the basic steps in REW were;
1) Create a Dirac impulse (on EQ screen, no filters applied, "create measurement from filters")
2) Select the Dirac impulse and apply bandpass filters to it, then "create measurement from predicted".
- For 33Hz target freq (the room length mode) I chose steep filters that maintained ~0dB at target freq, thus;
- - LowPass = 0.5 octave above (33*1.5=49.5), Butterworth, 48dB slope.
- - HighPass = 1 octave below (33/2=16.5), Butterworth, 48dB slope.
3) Create "Min phase version" on "Measurement actions" window (on "All SPL" screen).
4) Adjust the gain on the same "Measurement actions" window (add -6dB offset).
5) On "Impulse" screen, tick "Invert impulse".
6) Add delay of 60ms (click "Offset t=0" and add -60ms).
7) On "All SPL" screen use "Trace arithmetic" to add the unmodified Dirac impulse to the delayed impulse.
8) Open "IR windows" setting screen, and set Left=0, Ref=0, Right=200.
9) Export the IR (File > Export > Export IR as WAV). "Export measured IR", untick "Normalised", tick "Apply IR Window".

Now you have an IR to test with. In Audacity, create a 33Hz 1 cycle tone, and enough silence. Convolve it with the above IR. You will see the cancel tone isn't pure, but more like a wave packet. You will see the result does not have the correct amplitude, more like -7.2dBFS max. BUT, it possible the amplitude is "correct", because the "SPL" of the wave pack may very well be -6dB equivalent. Also the result is not delayed correctly, as you can see that the first strong valley is not at the right time compared to the manual example. So work out the difference and go back to REW and make the same adjustment, then export a new VBA. Then I convolved against the in room IR, and things were looking decent;


Now when I first saw the cancel wave convolved in Audacity and how it was now a "wave packet", I thought "that can't be good". Also in REW the SPL graph of the filter has the "comb" shape to it (some freq +dB, others -dB), but I wondered why? This IR has done nothing but add sound to the Dirac impulse (like an echo), so why has REW calc'd that SPL difference? Is it because some sounds "smear" the IR and increase the dB? But what about the -dB? So I also figured I better check the impulse against a bunch of other bass frequencies in Audacity. They all looked as expected - cancel tones of different amplitudes. I pushed ahead and convolved the results with the in room IR, and the results were better than expected! Here they are with +2dB amplification applied to all to make the swings more obvious;

The decay in every example is improved. Close to the target freq, ie 20/30/40Hz, are obviously better. 50/60/70Hz have reduced peak dB, but also "double +dB crests" are gone - they're back to a normal wave. 80Hz+ the improvements are minor, but somehow no peaks in the decay are worse than before!

Back in REW I convolved the new VBA with the old in room response, and the results were unremarkable;

But going to the spectrogram, here's the before and after;

Epic difference! I can't believe the decay improved so much. I don't know if real world performance will be the same? And this is just fixing the decay - still haven't looked at doing anything with the "build up" and "let down" of the main peak yet.

But I've also realised some more quirks of these IR, because when I use "Filtered IR" screen in REW, the result doesn't look the same. Similar, but not the same. And I worked out it's because the Filtered IR screen has bandpassed what it shows (the settings say as much - 1/1 or 1/3 octave), but the convolved results in Audacity are not bandpassed. If I bandpass the results in Audacity, things change! This means the in room IR takes a pure tone, and due to phase change etc, it happens to make other tones?! I wonder if that is how noise is captured in the IR? But also, the results in Audacity are inverted compared to REW. I'm not sure if that is relevant to this workflow?

Anyway, here are 33Hz in room responses I've shown before, but now bandpassed (LP=49.5, HP=16.5);

So it looks quite different when you remove the "noise", but still the main features are there.

So next I applied the same to 125Hz, because REW offers 1/1 filter for it;

So after bandpass and inverting in Audacity, I feel I'm looking at the same thing. I've highlighted some features that are in both, for confirmation.

As I write this I've had the thought that bandpass filters might be the cause of the "wave packet" looking result. So I bandpassed the 33Hz pure tone, and indeed it turns in to a sort of wave packet. So, I guess that's what that is. I'll have to investigate more. Because I was wondering if the "copy+delay+invert" method would be better because it will be "pure" (without wavepacket), but I haven't yet considered the bandpass on the copy.

Edit: Thought I better look at the 33Hz long tone test. I've been hesitant of OCA's VBA throughout this thread because the evidence seemed to suggest it would create a boosted first peak that could clip. Again that problem shows itself, but also it appears my new VBA has much better SPL without PEQ applied. In fact it seems my new VBA loses SPL after a few cycles. Truly impressive!

 

[OCA]

I tried a lot of possibilities and I've come up with some better VBA ideas. One of them especially excels among all others in terms of target tracking, clarity (not exactly minimizing group delay but flattest slope helping all frequencies sound right)

We all miss out the 2nd, and actually the largest (by a margin) response peak having focused on the first one all this time.

My initial idea was to combine the second and first peak VBA filters with each other. I tried all sorts of combinations. And although "dividing" (will be interesting to see @neRok try to divide impulses in Audacity!) 1st peak to the 2nd peak (keeping the 2nd peak impulse at normal polarity) gave the best "restoration" of the first dip among all other VBA filters, nothing was as good as simple VBA designed only on the second peak. No decrease in bass extension, a total of 7-8dB improvement in the first 2 dips (completely covering the second dip) and most importantly much better delay responses compared to others.

The division VBA "could" be useful for very low bass extension systems (@Keith_W?) with the first dip just after the audible frequency. The loss below 20Hz will not matter then and the first dip will be completely recovered.

The VBA filter applied responses below are after 20-200Hz auto EQ (no boost allowed) on psychoacoustic smoothed responses (including VBA free original response) for real life comparison:

 

[OCA]

Since I can't figure out how to not make FIR filters sound weird, minimum phase or not, I've gone back to good old parametric EQ!

If you share you measurements, I can have a look.

 

[Keith_W]

@OCA, I made a post in the Acourate forum since they have a thread about VBA's. Link here.

You might be interested to read Uli's reply. He proves that magnitude inversion creates a similar looking effect to the VBA. He does this by creating two filters - the first is an inversion of the bass response, and the second is a VBA computed from the first peak using the method you described (and in my post). He then tests the response of these two filters by convolving it with a sine wave and looking for differences. Result: the differences are very slight, and they can be considered equivalent.

I have proven this experimentally myself. Look at my plots comparing "VBA vs no VBA". The "No VBA" option that I used actually has Acourate's magnitude inversion built in. You can see that the result is extremely close - as I said in my post, "a sideways move".

I guess that if you believe Uli, that's the end of this thread! Nobody has to make VBA's any more! But it's still a cool thing to have And I learnt a lot by reverse engineering the room. As an exercise, it's a fun thing to do. But if you have Acourate and you apply Uli's suggested correction ... then maybe it's not needed.

 

[OCA]

@OCA, I made a post in the Acourate forum since they have a thread about VBA's. Link here.

You might be interested to read Uli's reply. He proves that magnitude inversion creates a similar looking effect to the VBA. He does this by creating two filters - the first is an inversion of the bass response, and the second is a VBA computed from the first peak using the method you described (and in my post). He then tests the response of these two filters by convolving it with a sine wave and looking for differences. Result: the differences are very slight, and they can be considered equivalent.

I have proven this experimentally myself. Look at my plots comparing "VBA vs no VBA". The "No VBA" option that I used actually has Acourate's magnitude inversion built in. You can see that the result is extremely close - as I said in my post, "a sideways move".

I guess that if you believe Uli, that's the end of this thread! Nobody has to make VBA's any more! But it's still a cool thing to have And I learnt a lot by reverse engineering the room. As an exercise, it's a fun thing to do. But if you have Acourate and you apply Uli's suggested correction ... then maybe it's not needed.

Thanks, I've read his comments. Inversion gives pretty good results up to 5dB allowed boost to dips (at least with the way it can be done in REW - Acourate might have a better tech for it), after that it will throttle the overall sound. And Uli has a point that this will easily compete if not beat 1D, room length based VBA.

But my recent experiments involving the other room dimensions are showing promise. The VBA above is sounding and measuring noticably better than what EQ can do. I was never comfortable with adding SPL offsets to the low pass filter and now I don't need to which is progress.

I am sure there's a 3D combination VBA for every room which will flatten the bass like magic but I am not sure if it will ever be possible to find it.

 

[ernestcarl]

I guess that if you believe Uli, that's the end of this thread! Nobody has to make VBA's any more! But it's still a cool thing to have And I learnt a lot by reverse engineering the room. As an exercise, it's a fun thing to do. But if you have Acourate and you apply Uli's suggested correction ... then maybe it's not needed.

I thought this was likely the case some time ago: https://www.audiosciencereview.com/.../preferred-target-fr-curve.39817/post-1533263

It would be nice to see more compelling actual before and after measurements. But I haven't seen much...

Dirac ART and Trinnov's "Waveforming" Bass Array technology are already so far ahead of this... more importantly, bass amplitude and decay becomes spatially more even/uniform across multiple seats.

And yet, even with the latter newer technology, you will still consistently hear from the makers of such DRC tools that some room acoustic treatment is still going to be desired for optimum performance.

 

[Tim Link]

If you share you measurements, I can have a look.

I figured it out. My bass horns are way louder than my little bookshelf speakers I'm using them with ( I know, weird) . I was taking advantage of that by just EQing the horns down to match the little speakers, which bought me super deep bass extension, measuring flat down to 5 Hz, Lol! What I wasn't catching was that the EQ was only lowering the pass band, but above the crossover it wasn't doing anything, so I was hearing out of passband in the horn relatively elevated. I pre-EQ'd the bass horns using parametric, making sure they followed the roll-off slope reasonably closely before using your REW impulse convlusion method for the listening position measurement. I also pre-EQ'd the bookshelf speakers to make them match very closely to each other, which improves the imaging.

Thanks again! Darn, I'm only getting down into the mid 20s now.

Turning down the horn's amp to be in the ballpark of the little speakers completely solved that problem. It sounds great! Definitely preferrable to what I can get with parametric EQ. That really deep notch at around 85 Hz seems to have something to do with the horns, although it might be a room interaction with the horns because I tested some little 10" subwoofers in the room as well and they had a decent notch at that frequency too. In any case it's narrow enough that I'm not noticing it as a serious problem.

 

[Tim Link]

And yet, even with the latter newer technology, you will still consistently hear from the makers of such DRC tools that some room acoustic treatment is still going to be desired for optimum performance.

I'm reminded of the quest for a way to determine longitude. Turned out they needed both an accurate time piece and accurate star and planet charts to allow ships to get an idea of what time it really was. The clock was only so accurate so needed to be adjusted now and then, but the stars were also only visible some of the time.

 

[neRok]

And yet, even with the latter newer technology, you will still consistently hear from the makers of such DRC tools that some room acoustic treatment is still going to be desired for optimum performance.

Which makes sense because we are trying to put cancel tones in to the room which are effectively more sounds in to the room that aren't part of the source material. So not having reflection related problems in the first place is definitely better than trying to actively cancel uncontrolled ones.

 

[Keith_W]

I thought this was likely the case some time ago: https://www.audiosciencereview.com/.../preferred-target-fr-curve.39817/post-1533263

It would be nice to see more compelling actual before and after measurements. But I haven't seen much...

If you look at my previous post, I compared "VBA" vs "no VBA" as the control. The VBA was computed from reverse engineering the room as per @OCA, the method was described in my post. It was a single reflection VBA taking into account one bass mode only. The "No VBA" actually had the bass prefilter described by Mitch in this video (time stamp 54:04):

This is the exact same method Uli recommends as a VBA alternative. So I was inadvertently replicating Uli's suggested method without realizing it. You can see in room measurements of my result - very little difference between the two.

Dirac ART and Trinnov's "Waveforming" Bass Array technology are already so far ahead of this... more importantly, bass amplitude and decay becomes spatially more even/uniform across multiple seats.

I am curious as to exactly what DSP Dirac and Trinnov are doing which is so different to a VBA. To my simple mind, it looks like they are using some kind of VBA, albeit in an automated fashion. Do you have any information about how their technology works?

 

[ernestcarl]

I am curious as to exactly what DSP Dirac and Trinnov are doing which is so different to a VBA. To my simple mind, it looks like they are using some kind of VBA, albeit in an automated fashion. Do you have any information about how their technology works?

What Dirac ART is using is a kind of multi-input multi-output (MIMO) EQ where the primary achievement is bass dereverberation using several (dynamically capable bass extended) speakers in more flexible multichannel configurations.

Trinnov "waveforming" is also a kind of MIMO approach but using (as far as we know) primarily front to back sub bass array arrangements.

Doing this type of extensive processing manually via REW (even with Acourate) etc. seems to me to be mere wishful thinking.

 

[neRok]

I tried a lot of possibilities and I've come up with some better VBA ideas. One of them especially excels among all others in terms of target tracking, clarity (not exactly minimizing group delay but flattest slope helping all frequencies sound right)

We all miss out the 2nd, and actually the largest (by a margin) response peak having focused on the first one all this time.

So after you posted this I realised I hadn't looked at your measurements from a "classic" EQ'ing point of view. And actually your old Left measure has some weird things going on, which I had noticed before, but hadn't compared to the right. Now I see how bad and odd the issues are. The following image highlights the strange things in the left measure;


So was anything substantially different about where you had the left speaker located? You mentioned an opening on the back wall, so perhaps the backwave reflection is only affecting 1 speaker? I wonder if moving the mic slightly and doing a new measure would have "tricked" REW in to handling the timing right around that 50Hz problem?

Because with how REW has decided the strong delayed wave at ~52Hz is the "main wave", I am concerned how this becomes the timing reference point, and how the phase at the moment is treated as the "Phase" on the phase plots etc. Because in the next screenshot I simply added up your Left and Right measures. Ideally the sum should be +6dB at all points (and a stereo sweep should be similar to the sum). Anytime the sum dips below below one of the measurements indicates they are out of phase at that frequency at the listening position. And the screenshot below shows obvious phase difference between 40Hz and 60Hz.


But again, is it a true phase issue, or an oddity with where REW has referenced the phase based upon the SPL peak wave?

I mention the above because it is probably best to fix any such phase issue before trying to lay VBA fix over top. Does your new room layout have any such issues?

On the topic of phase issues, I finally took a look at my own measurements. Now that I have a better understanding, they don't look so bad (or as mysterious). I determined a phase issue in my measurements at ~93Hz, because the combined measurement dropped below the individual ones;

Also in that measurement I have determined all the peaks I would hope to address with VBA = the red numbers. The 2 green peaks are because my Left speaker's direct wave goes in to an odd shaped corner behind me, so it comes back as a "second back wave".
But on the phase issue, I worked an all pass on the Left speaker would fix that (and a PEQ for the ~160Hz peak).

And then a new measurement confirmed the fix;

The above was all moving mic measurements, so I did 2 sweeps to be sure;

The VBA filter applied responses below are after 20-200Hz auto EQ (no boost allowed) on psychoacoustic smoothed responses (including VBA free original response) for real life comparison:

Psychoacoustic smoothing is kind of cheating lol. It makes sense when not using any sort of "bass array" (or active room treatment), because narrow Q peaks and nulls aren't that relevant when considering the whole sound spectrum or even just bass region. fineMen eluded to as much in their previous post: "Let there be some peaks and dips, because the human hearing is a tolerant, if not to say sloppy, in this range".

I think otherwise. REW seems to only show the result of a 1 cycle tone in the room. But like I've shown a few times, a room mode will build up power over the first few cycles, until it "settles" at a new maximum. REW does not directly show this phenomenon in its measurements, but it is implied on the timing graphs (because the reflections are captured in the IR). And this isn't something I've made up, because this is what the Wiki page on Room modes says: These standing waves can be considered a temporary storage of acoustic energy as they take a finite time to build up and a finite time to dissipate once the sound energy source has been removed.

 

[Keith_W]

What Dirac ART is using is a kind of multi-input multi-output (MIMO) EQ where the primary achievement is bass dereverberation using several (dynamically capable bass extended) speakers in more flexible multichannel configurations.

Trinnov "waveforming" is also a kind of MIMO approach but using (as far as we know) primarily front to back sub bass array arrangements.

Doing this type of extensive processing manually via REW (even with Acourate) etc. seems to me to be mere wishful thinking.

Thank you for your response. I am still trying to wrap my head around that paper you linked to about MIMO. It is quite hard going to read an academic paper when you only have high school maths which you've mostly forgotten!

I watched that Trinnov video. Twice. It is excellent and I have sent links to it to some of my friends, so thank you very much for that! However, I don't see anything different to a VBA. He also mentions bass steering, which I think is also within the ability of Acourate and (maybe REW?) to accomplish. So if I missed something important, I would appreciate if you could tell me in what way the Trinnov correction is different to a VBA?

 

[ernestcarl]

I am still trying to wrap my head around that paper you linked to about MIMO. It is quite hard going to read an academic paper when you only have high school maths which you've mostly forgotten!

I don't think you need advanced math degrees to get the "gist" of MIMO equalization... but it certainly helps some if you read past academic papers that delve on the subject matter and try hard enough to understand at the very least why a very high level of control over multiple distributed LF transducers in small room acoustics matters.

I don't see anything different to a VBA. He also mentions bass steering, which I think is also within the ability of Acourate and (maybe REW?) to accomplish. So if I missed something important, I would appreciate if you could tell me in what way the Trinnov correction is different to a VBA?

Yes, I think what sets apart Trinnov waveforming is the planar array bass steering, where -- as they say -- up to 90% of the LF energy is already fully absorbed by the rear subs alone. It also looks like it will work really well for larger home cinema spaces.

VBA at it's simplest form using a single-input single output (SISO) configuration isn't capable of solving bass spatial amplitude variability and decay reverberation. While it is, of course, attempting to EQ compensate or "cancel" out the biggest initial LF peak and dip caused by the room's length and/or width modes, it does it very poorly (based on the limited real in-room measurements I've seen) and this has been only thus far shown from a single microphone position.

Tell me, honestly, does the amount of bass "cancellation" achieved below by this inverse, time delayed VBA filter look at all that significant to you?

 

[Hayabusa]

I don't think you need advanced math degrees to get the "gist" of MIMO equalization... but it certainly helps some if you read past academic papers that delve on the subject matter and try hard enough to understand at the very least why a very high level of control over multiple distributed LF transducers in small room acoustics matters.



Yes, I think what sets apart Trinnov waveforming is the planar array bass steering, where -- as they say -- up to 90% of the LF energy is already fully absorbed by the rear subs alone. It also looks like it will work really well for larger home cinema spaces.

VBA at it's simplest form using a single-input single output (SISO) configuration isn't capable of solving bass spatial amplitude variability and decay reverberation. While it is, of course, attempting to EQ compensate or "cancel" out the biggest initial LF peak and dip caused by the room's length and/or width modes, it does it very poorly (based on the limited real in-room measurements I've seen) and this has been only thus far shown from a single microphone position.

Tell me, honestly, does the amount of bass "cancellation" achieved below by this inverse, time delayed VBA filter look at all that significant to you?

Where do I see the result graph to judge if it is significant?